Reliable CPS Design

Traditionally, the reliability of safety-critical systems has been estimated via rigorous testing. However, testing a large system is cumbersome, costly, and even infeasible sometimes. Additionally, if testing deems a system to be unreliable as per the required standard, it is hard to redesign a complete system. As an alternative, we advocate the use of formal methods to verify system reliability in its early stages of design using its behavioral description. We also propose suitable design modifications to adhere the system to its required reliability standard.



Research Topics

# Reliability Analysis of CPS

A reliability specification language RELSPEC is developed, which models component-based systems using a Discrete Time Markov Chain and estimates its reliability against permanent hardware faults using probabilistic model checking. We have also proposed suitable design space exploration methods toward designing a reliable system given the reliability options available for different components.

relsys

# Reliability-Aware Scheduling of CPS

By verifying the performance degradation of a control system against transient sensory faults, suitable sensor fault mitigation techniques are developed. The system shows significant reliability improvement using developed techniques. We have also proposed timed automata based strategies for scheduling periodic control tasks with sporadic sensor fault mitigation on shared multiprocessor architectures. Thus reliability-aware scheduling strategies for periodic and sporadic real-time tasks on heterogeneous embedded systems are developed. In future, we plan to explore clustering techniques to optimize the table size used in these developed scheduling strategies.

relsched